OpenAI secures new partnership amidst Asia expansion
Up next
Author
Tags

OpenAI has introduced GPT-4.5, internally referred to as ‘Orion,’ marking its most advanced AI language model to date. GPT-4.5 offers enhanced writing capabilities, deeper world knowledge, and a refined conversational style, making interactions more natural and effective for tasks like writing, programming, and problem-solving.

OpenAI founder Sam Altman has termed the model more “thoughtful” and something that has felt more natural. “good news: it is the first model that feels like talking to a thoughtful person to me. i have had several moments where i’ve sat back in my chair and been astonished at getting actually good advice from an AI”, he wrote on X.

In technicality though, GPT-4.5 serves as a transitional model, bridging the gap between traditional AI systems and forthcoming models with advanced reasoning capabilities. GPT-4.5 offers enhanced computational efficiency, boasting over 10 times the efficiency of GPT-4, leading to faster and more reliable responses.

“GPT‑4.5 has a better understanding of what humans mean and interprets subtle cues or implicit expectations with greater nuance and “EQ”. GPT‑4.5 also shows stronger aesthetic intuition and creativity. It excels at helping with writing and design,” the company said in its statement.

In terms of accuracy, the model demonstrates a significant reduction in generating inaccurate information, with hallucination rates decreasing to 37% from nearly 60% observed in its predecessor, GPT-4o. This might be because the model has been trained with new supervision techniques, combined with traditional supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF).

“GPT‑4.5 doesn’t think before it responds, which makes its strengths particularly different from reasoning models like OpenAI o1. Compared to OpenAI o1 and OpenAI o3‑mini, GPT‑4.5 is a more general-purpose, innately smarter model,” the ChatGPT maker said.

However, this new model comes with certain limitations, at least for now. Currently, GPT-4.5 does not support multimodal capabilities like Voice Mode, video, or screen sharing within ChatGPT. Additionally, due to its size and computational demands, GPT-4.5 is comparatively more expensive to operate and is not intended to replace GPT-4o.

Speaking of availability, for now, GPT-4.5 is available as a research preview, allowing users to explore its capabilities and provide feedback for future improvements. GPT-4.5 is available via the Chat Completions API, Assistants API, and Batch API, supporting features like function calling, structured outputs, streaming, system messages, and image inputs.

In fact, users who pay $200 a month for ChatGPT Pro will get access to GPT-4.5 first, starting Thursday, while developers using OpenAI’s paid API can start using GPT-4.5 today (February 28). In the meantime, other paid users, like those with ChatGPT Plus ($20/month) and ChatGPT Team, will get access next week.

This release reflects OpenAI’s commitment to maintaining its leadership in the competitive AI landscape. But OpenAI’s roadmap hinted that the upcoming GPT-5 will integrate various other technologies, including the o3 reasoning model, aiming to create a more unified and versatile AI system.

As global competition increases with opponents like DeepSeek, OpenAI is also focusing on AI agents, including its recent launch of Operator. On the financial side, the AI trendsetter has raised a total of $17.9 billion in funding over 10 rounds. The Sam Altman-led company reached a valuation of approximately $157 billion in its last funding round in October 2024.

0
0
0